huggingface/candle
1
0
Fork 0
mirror of https://github.com/huggingface/candle.git synced 2025-06-17 02:58:50 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fix for parler-tts, do not add the last slice of padding tokens. (#2442)

Browse Source 
* Fix for parler-tts, do not add the last slice of padding tokens.

* Support for the mini model.
This commit is contained in:
Laurent Mazare
2024-08-22 22:22:03 +01:00
committed by GitHub
parent e3c146ada6
commit 2ec8729d51
2 changed files with 21 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
candle-examples/examples/parler-tts/main.rs
View File

@ -86,6 +86,17 @@ struct Args {
/// The output wav file. /// The output wav file.
#[arg(long, default_value = "out.wav")] #[arg(long, default_value = "out.wav")]
out_file: String, out_file: String,
#[arg(long, default_value = "large-v1")]
which: Which,
}
#[derive(Clone, Debug, Copy, PartialEq, Eq, clap::ValueEnum)]
enum Which {
#[value(name = "large-v1")]
LargeV1,
#[value(name = "mini-v1")]
MiniV1,
} }
fn main() -> anyhow::Result<()> { fn main() -> anyhow::Result<()> {
@ -117,7 +128,10 @@ fn main() -> anyhow::Result<()> {
let api = hf_hub::api::sync::Api::new()?; let api = hf_hub::api::sync::Api::new()?;
let model_id = match args.model_id { let model_id = match args.model_id {
Some(model_id) => model_id.to_string(), Some(model_id) => model_id.to_string(),
None => "parler-tts/parler-tts-large-v1".to_string(), None => match args.which {
Which::LargeV1 => "parler-tts/parler-tts-large-v1".to_string(),
Which::MiniV1 => "parler-tts/parler-tts-mini-v1".to_string(),
},
}; };
let revision = match args.revision { let revision = match args.revision {
Some(r) => r, Some(r) => r,
@ -130,7 +144,12 @@ fn main() -> anyhow::Result<()> {
)); ));
let model_files = match args.model_file { let model_files = match args.model_file {
Some(m) => vec![m.into()], Some(m) => vec![m.into()],
None => candle_examples::hub_load_safetensors(&repo, "model.safetensors.index.json")?, None => match args.which {
Which::MiniV1 => vec![repo.get("model.safetensors")?],
Which::LargeV1 => {
candle_examples::hub_load_safetensors(&repo, "model.safetensors.index.json")?
}
},
}; };
let config = match args.config_file { let config = match args.config_file {
Some(m) => m.into(), Some(m) => m.into(),

1
candle-transformers/src/models/parler_tts.rs
View File

@ -429,7 +429,6 @@ impl Model {
let min_len = all_audio_tokens.iter().map(|v| v.len()).min().unwrap_or(0); let min_len = all_audio_tokens.iter().map(|v| v.len()).min().unwrap_or(0);
all_audio_tokens.iter_mut().for_each(|v| { all_audio_tokens.iter_mut().for_each(|v| {
v.resize(min_len, 0); v.resize(min_len, 0);
v.push(self.pad_token_id)
}); });
let all_audio_tokens = Tensor::new(all_audio_tokens, &candle::Device::Cpu)?; let all_audio_tokens = Tensor::new(all_audio_tokens, &candle::Device::Cpu)?;
Ok(all_audio_tokens) Ok(all_audio_tokens)